p8105_final_jmmm

#project website. can you see it? https://josiealford14.github.io/therealcostofdrugs/

library(tidyverse)
library(dplyr)
library(rvest)
library(httr)
library(janitor)
library(stringr)
library(forcats)
library(tidytext)
library(viridis)
library(broom)
library(plotly)

Dataset 1: Drug Prices

• Looking at the National Drug Acquisition Cost dataset:

# # API code
# url = "https://data.medicaid.gov/resource/tau9-gfwr.csv"
# drug_price =
#   GET(url, query = list("$limit" = 9999999)) %>%
#   content("parsed") %>%
#   clean_names()

# # CSV code
drug_price =
  read_csv("../data/drug_price.csv") %>%
  clean_names()

## Warning: Missing column names filled in: 'X1' [1]

## Parsed with column specification:
## cols(
##   X1 = col_integer(),
##   as_of_date = col_date(format = ""),
##   classification_for_rate_setting = col_character(),
##   corresponding_generic_drug_effective_date = col_character(),
##   corresponding_generic_drug_nadac_per_unit = col_character(),
##   effective_date = col_date(format = ""),
##   explanation_code = col_character(),
##   nadac_per_unit = col_double(),
##   ndc = col_character(),
##   ndc_description = col_character(),
##   otc = col_character(),
##   pharmacy_type_indicator = col_character(),
##   pricing_unit = col_character()
## )

• The raw drug_price dataset is huge and takes a while to load.
• It contains weekly drug prices from november 2013 to october 2017 recorded. There are 4902795 observations and 13 variables in the dataset:
  • as_of_date: date of data posting
  • classification_for_rate_setting: indicates whether, for NADAC calculation, the NDC was considered brand (‘B’) or generic (‘G’) or brand (‘B’) approved under an Abbreviated New Drug Application (ANDA) (‘B-ANDA’).
  • corresponding_generic_drug_effecitve_date: effective date of when the Corresponding Generic Drug NADAC Per Unit is assigned to a multiple source brand drug NDC
  • corresponding_generic_drug_nadac_per_unit: NADAC for the corresponding generic drug
  • effective_date: effective date of the NADAC Per Unit cost
  • explanation_code: codes that pertain to how the NADAC was calculated. see explanation code descriptions here
  • nadac_per_unit: National Average Drug Acquisition Cost per unit
  • ndc: 11-digit FDA code that includes the labeler code, product code, and package code
  • ndc_description: drug name, strength, and dosage form of the drug
  • otc: over-the-counter (OTC) product (‘Y’ or ‘N’)
  • pharmacy_type_indicator: source of pharmacy survey data used to calculate the NADAC. ‘C/I’ indicates data was collected from surveys of Chain/Independent pharmacies. Other pharmacy type indicators are not used at this time.
  • pricing_unit: pricing unit for the associated NDC (‘ML’, ‘GM’ or ‘EA’)
• We will focus on the variables effective_date, ndc, ndc_description, otc, classification_for_rate_setting, pricing_unit, and nadac_per_unit. We will rename the variables effective_date, ndc_description, classification_for_rate_setting, and nadac_per_unit as date, drug_name, drug_type, and unit_price respectively. We will also split the date variable into the year, month and date variables.

drug_price = 
  drug_price %>%
  rename(trade_name = ndc_description,
         drug_type = classification_for_rate_setting,
         unit_price = nadac_per_unit) %>%
  mutate(trade_name = str_replace(trade_name, "[0-9]", "_"),
         trade_name = str_to_title(trade_name),
         otc = recode(otc, "Y" = "yes", "N" = "no"),
         otc = str_to_title(otc),
         otc = as.factor(otc)) %>% 
  separate(trade_name, into = c("trade_name", "remove"), sep = "_") %>% 
  separate(effective_date, into = c("year", "month", "date")) %>%
  select(-c(date, x1, remove)) %>%
  mutate(trade_name = trimws(trade_name, which = c("both"))) %>% 
  separate(trade_name, into = c("trade_name", "remove"), sep = " ") %>% 
  filter(is.na(remove)) %>%
  select(-remove) %>% 
  filter(pricing_unit == "GM" | pricing_unit == "EA") %>% 
  select(trade_name, everything()) 

## Warning: Too few values at 308308 locations: 1, 2, 7, 17, 18, 19, 30, 45,
## 48, 56, 59, 60, 64, 66, 68, 85, 86, 87, 92, 107, ...

## Warning in format.POSIXlt(as.POSIXlt(x), ...): unknown timezone 'zone/tz/
## 2017c.1.0/zoneinfo/America/New_York'

## Warning: Too many values at 592722 locations: 5, 17, 18, 48, 68, 112, 128,
## 136, 138, 173, 196, 215, 222, 231, 235, 241, 265, 273, 279, 294, ...

## Warning: Too few values at 2730099 locations: 1, 3, 4, 6, 8, 9, 10, 11, 12,
## 13, 15, 16, 20, 21, 23, 24, 25, 26, 27, 28, ...

• We look at the yearly distribution of drug prices for a few drugs. Some drugs have skewed price distributions and the median (instead of the mean) price might be the better measure to use in the analyses. Also, it might be worth creating a shiny app for people to view the price distributions for various drugs.

drug_price %>%
  filter(trade_name == "Abacavir") %>%
  ggplot(aes(x = year, y = unit_price)) + geom_violin(aes(fill = year), color = "blue", alpha = .5)

new_drug_price = drug_price %>% 
  group_by(trade_name, drug_type, otc, pricing_unit, year) %>% 
  summarize(avg_price = mean(unit_price, na.rm = TRUE))

new_drug_price

## # A tibble: 6,023 x 6
## # Groups:   trade_name, drug_type, otc, pricing_unit [?]
##    trade_name drug_type    otc pricing_unit  year avg_price
##         <chr>     <chr> <fctr>        <chr> <chr>     <dbl>
##  1                    G    Yes           EA  2013 0.2434375
##  2                    G    Yes           EA  2014 0.2430778
##  3                    G    Yes           EA  2015 0.2556137
##  4                    G    Yes           EA  2016 0.2538263
##  5                    G    Yes           EA  2017 0.2339103
##  6                    G    Yes           GM  2016 0.3077704
##  7                    G    Yes           GM  2017 0.3087167
##  8   Abacavir         G     No           EA  2013 5.3170850
##  9   Abacavir         G     No           EA  2014 4.7696039
## 10   Abacavir         G     No           EA  2015 3.6822746
## # ... with 6,013 more rows

• Points to consider:
  • Some unit prices are listed as price per grams or price per milliliters or price per tablet We need to create a price variable with the appropriate unit price if we want to compare drug prices. One way to do this is to convert all unit prices to price per grams by dividing the price of each tablet by the tablet weight and by using the approximate relationship 1ml = 1gm.

Dataset 2: Funding for Research

• The NIH RCDC Report
• This dataset contains the annual support for various disease categories based on grants, contracts, and other funding mechanisms used across the National Institutes of Health (NIH). Prevalence and mortality data included in this dataset is from the National Center for Health Statistics (NCHS) at the Centers for Disease Control & Prevention (CDC). *Funding for 2018 are estimated figures.

url = "https://report.nih.gov/categorical_spending.aspx"
NIH_xml = read_html(url)
NIH_funding = (NIH_xml %>% html_nodes(css = "table"))[[3]] %>% html_table() %>% as_tibble()
NIH_funding = NIH_funding %>%
  clean_names() %>%
  mutate(US_mortality_2015 = x2015_us_mortality_19)

NIH_funding$x2015_us_prevalence_standard_error_19  = recode(NIH_funding$x2015_us_prevalence_standard_error_19 , "-" = "NA")
NIH_funding$US_mortality_2015 = recode(NIH_funding$US_mortality_2015, "-" = "NA")

NIH_funding = NIH_funding %>%
  separate(x2015_us_prevalence_standard_error_19, into = c("prevalence", "remove"), sep = "\\%") %>%
  select(-x2015_us_mortality_19, -remove)

## Warning: Too many values at 36 locations: 7, 19, 21, 24, 25, 35, 37, 40,
## 46, 48, 54, 59, 74, 80, 81, 89, 93, 114, 115, 117, ...

## Warning: Too few values at 246 locations: 1, 2, 3, 4, 5, 6, 8, 9, 10, 11,
## 12, 13, 14, 15, 16, 17, 18, 20, 22, 23, ...

NIH_funding$fy_2013actual = recode(NIH_funding$fy_2013actual, "+" = "NNA")
NIH_funding$fy_2014actual = recode(NIH_funding$fy_2014actual, "+" = "NNA")
NIH_funding$fy_2015actual = recode(NIH_funding$fy_2015actual, "+" = "NNA")
NIH_funding$fy_2016actual = recode(NIH_funding$fy_2016actual, "+" = "NNA")

NIH_funding = NIH_funding %>%
  mutate(fy_2013 = as.numeric(substring(fy_2013actual, 2))) %>%
  mutate(fy_2014 = as.numeric(substring(fy_2014actual, 2))) %>%
  mutate(fy_2015 = as.numeric(substring(fy_2015actual, 2))) %>%
  mutate(fy_2016 = as.numeric(substring(fy_2016actual, 2))) %>%
  mutate(fy_2017 = as.numeric(substring(fy_2017estimated_enacted, 2))) %>%
  mutate(fy_2018estimated = as.numeric(substring(fy_2018estimated, 2))) 

## Warning in evalq(as.numeric(substring(fy_2013actual, 2)), <environment>):
## NAs introduced by coercion

## Warning in evalq(as.numeric(substring(fy_2014actual, 2)), <environment>):
## NAs introduced by coercion

## Warning in evalq(as.numeric(substring(fy_2015actual, 2)), <environment>):
## NAs introduced by coercion

## Warning in evalq(as.numeric(substring(fy_2016actual, 2)), <environment>):
## NAs introduced by coercion

## Warning in evalq(as.numeric(substring(fy_2017estimated_enacted, 2)),
## <environment>): NAs introduced by coercion

## Warning in evalq(as.numeric(substring(fy_2018estimated, 2)),
## <environment>): NAs introduced by coercion

NIH_funding$prevalence = as.numeric(NIH_funding$prevalence)

## Warning: NAs introduced by coercion

NIH_funding$US_mortality_2015 = as.numeric(NIH_funding$US_mortality_2015)

## Warning: NAs introduced by coercion

NIH_funding = NIH_funding %>%
  select(-c(fy_2013actual, fy_2014actual, fy_2015actual, fy_2016actual, fy_2017estimated_enacted))

final_NIH_funding = NIH_funding %>%
  rename(disease = research_disease_areas_dollars_in_millions_and_rounded, 
         '2013' = fy_2013,
         '2014' = fy_2014,
         '2015' = fy_2015,
         '2016' = fy_2016,
         '2017' = fy_2017,
         '2018' = fy_2018estimated) %>%
  gather(key = "year", value = funding, c("2013", "2014", "2015", "2016", "2017", "2018")) %>% 
  separate(disease, into = c("disease", "remove"), sep = " ") %>% 
  select(-remove)

## Warning: Too many values at 666 locations: 1, 2, 3, 4, 6, 7, 10, 11, 12,
## 24, 28, 30, 38, 50, 52, 53, 54, 55, 56, 58, ...

## Warning: Too few values at 384 locations: 5, 8, 13, 14, 17, 18, 19, 21, 23,
## 25, 32, 33, 39, 40, 44, 47, 66, 71, 73, 84, ...

rm(NIH_funding)

final_NIH_funding

## # A tibble: 1,692 x 5
##        disease prevalence US_mortality_2015  year funding
##  *       <chr>      <dbl>             <dbl> <chr>   <dbl>
##  1    Acquired         NA                NA  2013      NA
##  2       Acute         NA                NA  2013      95
##  3  Adolescent         NA               138  2013      70
##  4       Agent         NA                NA  2013      10
##  5       Aging         NA                NA  2013      NA
##  6 Alcoholism,         NA                NA  2013     437
##  7    Allergic        8.2                77  2013       9
##  8         ALS         NA                NA  2013      39
##  9 Alzheimer's         NA                NA  2013     504
## 10 Alzheimer's         NA                NA  2013      NA
## # ... with 1,682 more rows

Dataset 3: Patents

• The Patents dataset from the FDA Orange Book

• Below is a description of the dataset and variables from https://www.fda.gov/Drugs/InformationOnDrugs/ucm129689.htm

• New Drug Application Type: The type of new drug application approval, either New Drug Applications(NDA or innovator) or Abbreviated New Drug Applications(ANDA or generic)

• New Drug Application (NDA) Number (appl_no) : The FDA assigned number to the application.

• Product Number: The FDA assigned number to identify the application products. Each strength is a separate product. May repeat for multiple part products.

• Patent Number (patent_num): Patent numbers as submitted by the applicant holder for patents covered by the statutory provisions. May repeat for multiple applications and multiple products.

• Pediatric Exclusivity Granted: Yes or No. Was pediatric exclusivity granted by the agency?

• Patent Expire Date: The date the patent expires as submitted by the applicant holder including applicable extensions. This was split into Year, Month, and Day variables

• Drug Substance Flag: Patents submitted on FDA Form 3542 and listed after August 18, 2003 may have a drug substance flag indicating the sponsor submitted the patent as claiming the drug substance.

• Drug Product Flag: Patents submitted on FDA Form 3542 and listed after August 18, 2003 may have a drug product flag indicating the sponsor submitted the patent as claiming the drug product.

• Patent Use Code: Code to designate a use patent that covers the approved indication or use of a drug product. May repeat for multiple applications, multiple products and multiple patents.

• Patent Delist Request Flag: Sponsor has requested patent be delisted. This patent has remained listed because, under Section 505(j)(5)(D)(i) of the Act, a first applicant may retain eligibility for 180-day exclusivity based on a paragraph IV certification to this patent for a certain period. Applicants under Section 505(b)(2) are not required to certify to patents where this flag is set to Y.

patents = read_csv("./data/patent.csv")

## Parsed with column specification:
## cols(
##   appl_type = col_character(),
##   appl_no = col_integer(),
##   product_no = col_integer(),
##   patent_no = col_character(),
##   patexpstr = col_character(),
##   patexp = col_character(),
##   drugsub = col_character(),
##   drugprod = col_character(),
##   patuse = col_character(),
##   delist = col_character()
## )

patents = patents %>% 
  clean_names() %>%
  rename(drug_type = appl_type) %>% 
  mutate(drug_type = recode(drug_type, "N" = "B")) %>%
  separate(patexp, into = c("month_pat_expires", "day_pat_expires", "year_pat_expires"), sep = "/") %>%
  separate(patent_no, into = c("patent_num", "pediatric_exclusivity_granted"), sep = "\\*")

## Warning: Too few values at 12167 locations: 1, 2, 3, 5, 7, 9, 10, 11, 12,
## 13, 16, 17, 18, 19, 20, 21, 22, 23, 24, 26, ...

patents$pediatric_exclusivity_granted[is.na(patents$pediatric_exclusivity_granted)] = "no"
patents$drugsub[is.na(patents$drugsub)] = "no"
patents$drugprod[is.na(patents$drugprod)] = "no"
patents$delist[is.na(patents$delist)] = "no"

patents = patents %>%
  mutate(pediatric_exclusivity_granted = factor(pediatric_exclusivity_granted, levels = c("no", "PED"), labels = c("No", "Yes"))) %>%
  mutate(drug_substance_flag = factor(drugsub, levels = c("no", "Y"), labels = c("no", "yes"))) %>%
  mutate(drug_prod_flag = factor(drugprod, levels = c("no", "Y"), labels = c("no", "yes"))) %>%
  mutate(patent_delisted = factor(delist, levels = c("no", "Y"), labels = c("no", "yes"))) %>%
  select(-drugsub, -drugprod, -patexpstr) %>%
  rename(pat_use_code = patuse)
patents

## # A tibble: 13,858 x 13
##    drug_type appl_no product_no patent_num pediatric_exclusivity_granted
##        <chr>   <int>      <int>      <chr>                        <fctr>
##  1         B   22425          1    5223510                            No
##  2         B   22425          1    5223510                            No
##  3         B   22307          1    5288726                            No
##  4         B   22307          1    5288726                           Yes
##  5         B   21462          1    5344932                            No
##  6         B   21462          1    5344932                           Yes
##  7         B   22088          1    5362718                            No
##  8         B   22088          1    5362718                           Yes
##  9         B   21060          1    5364842                            No
## 10         B   21060          1    5364842                            No
## # ... with 13,848 more rows, and 8 more variables:
## #   month_pat_expires <chr>, day_pat_expires <chr>,
## #   year_pat_expires <chr>, pat_use_code <chr>, delist <chr>,
## #   drug_substance_flag <fctr>, drug_prod_flag <fctr>,
## #   patent_delisted <fctr>

Dataset 4: Products

*The Products dataset from the FDA Orange Book

• Below is a description of the dataset and variables from https://www.fda.gov/Drugs/InformationOnDrugs/ucm129689.htm

• Ingredient: The active ingredient(s) for the product. Multiple ingredients are in alphabetical order, separated by a semicolon.

• Dose form;The product dosage form

• Route: Product dosage route

• Trade_Name: The trade name of the product as shown on the labeling.

• Applicant: The firm name holding legal responsibility for the new drug application. The firm name is condensed to a maximum twenty character unique string.

• Strength: The potency of the active ingredient. May repeat for multiple part products.

• Drug Application Type: The type of new drug application approval. NDA orANDA as for patents dataset.

• New Drug Application (NDA) Number: The FDA assigned number to the application.

• Product Number: The FDA assigned number to identify the application products. Each strength is a separate product. May repeat for multiple part products.

• Therapeutic Equivalence (TE) Code: The TE Code indicates the therapeutic equivalence rating of generic to innovator Rx products.

• Approval Date: The date the product was approved as stated in the FDA approval letter to the applicant. Products approved prior to the January 1, 1982 contain the phrase: “Approved prior to Jan 1, 1982”. This separated into Year, Month, and Day variables.

• Reference Listed Drug (RLD) Updated: The RLD is a drug product approved under section 505(c) of the FD&C Act for which FDA has made a finding of safety and effectiveness. In the electronic Orange Book, an RLD is identified by “RLD” in the RLD column.

• Reference Standard (RS) New! :A “reference standard” is the drug product selected by FDA that an applicant seeking approval of an ANDA must use in conducting an in vivo bioequivalence study required for approval of an ANDA. In the electronic Orange Book, a reference standard is identified by “RS” in the RS column.

*Type: The group or category of approved drugs. RX (prescription), OTC (over the counter), or DISCN (discontinued).

*Applicant Full Name: The full name of the firm holding legal responsibility for the new drug application.

products = read_csv("./data/products.csv")

## Parsed with column specification:
## cols(
##   ingredient = col_character(),
##   dfroute = col_character(),
##   doseform = col_character(),
##   route = col_character(),
##   trade_name = col_character(),
##   applicant = col_character(),
##   strength = col_character(),
##   appl_type = col_character(),
##   appl_no = col_integer(),
##   product_no = col_integer(),
##   tecodestr = col_character(),
##   tecode = col_integer(),
##   appdatestr = col_character(),
##   appdate = col_character(),
##   rld = col_character(),
##   rs = col_character(),
##   type = col_character(),
##   appfull = col_character()
## )

products = products %>%
  rename(drug_type = appl_type,
         otc = type) %>% 
  mutate(drug_type = recode(drug_type, "A" = "G", "N" = "B"),
         otc = recode(otc, "OTC" = "yes", "RX" = "no", "DISCN" = "no")) %>%
  separate(appdate, into = c("approval_month", "approval_day", "approval_year"), sep = "/") %>%
  #separate(doseform, into = c("form", "release_type"), sep = ",")
  mutate(otc = factor(otc)) %>%
  mutate(ref_listed_drug = factor(rld)) %>%
  mutate(ref_std = factor(rs)) %>%
  select(-appdatestr, -rld, -rs, -dfroute)
products

## # A tibble: 34,209 x 18
##                                           ingredient doseform route
##                                                <chr>    <chr> <chr>
##  1                                  ABACAVIR SULFATE SOLUTION  ORAL
##  2                                  ABACAVIR SULFATE SOLUTION  ORAL
##  3                                  ABACAVIR SULFATE   TABLET  ORAL
##  4                                  ABACAVIR SULFATE   TABLET  ORAL
##  5                                  ABACAVIR SULFATE   TABLET  ORAL
##  6                                  ABACAVIR SULFATE   TABLET  ORAL
##  7                                  ABACAVIR SULFATE   TABLET  ORAL
##  8                                  ABACAVIR SULFATE   TABLET  ORAL
##  9 ABACAVIR SULFATE; DOLUTEGRAVIR SODIUM; LAMIVUDINE   TABLET  ORAL
## 10                      ABACAVIR SULFATE; LAMIVUDINE   TABLET  ORAL
## # ... with 34,199 more rows, and 15 more variables: trade_name <chr>,
## #   applicant <chr>, strength <chr>, drug_type <chr>, appl_no <int>,
## #   product_no <int>, tecodestr <chr>, tecode <int>, approval_month <chr>,
## #   approval_day <chr>, approval_year <chr>, otc <fctr>, appfull <chr>,
## #   ref_listed_drug <fctr>, ref_std <fctr>

Dataset 5: Exclusivity

We load and tidy the exclusivity dataset

exclusivity = read_csv("./data/exclusivity.csv") %>%
  clean_names() %>%
  rename(drug_type = appl_type) %>% 
  mutate(drug_type = recode(drug_type, "A" = "G", "N" = "B")) %>% 
  separate(excldate, into = c("month_excl_expires", "date_excl_expires", "year_excl_expires"))

## Parsed with column specification:
## cols(
##   appl_type = col_character(),
##   appl_no = col_integer(),
##   product_no = col_integer(),
##   exclcode = col_character(),
##   excldatestr = col_character(),
##   excldate = col_character()
## )

exclusivity

## # A tibble: 1,806 x 8
##    drug_type appl_no product_no exclcode  excldatestr month_excl_expires
##  *     <chr>   <int>      <int>    <chr>        <chr>              <chr>
##  1         G   65488          1       PC Mar 26, 2018                  3
##  2         G   65488          2       PC Mar 26, 2018                  3
##  3         G   78276          1       PC Apr 24, 2017                  4
##  4         G   78276          2       PC Apr 24, 2017                  4
##  5         G   78276          3       PC Apr 24, 2017                  4
##  6         G   78340          1       PC Jul 30, 2016                  7
##  7         G   78340          2       PC Jul 30, 2016                  7
##  8         G   78560          1       PC Jun 10, 2017                  6
##  9         G   78827          1       PC Apr 24, 2017                  4
## 10         G   78827          2       PC Apr 24, 2017                  4
## # ... with 1,796 more rows, and 2 more variables: date_excl_expires <chr>,
## #   year_excl_expires <chr>

The tidied exclusivity dataset has 1806 observations and 8 variables:

• year, month, and date the exclusivity expires
• appl_type: type of new drug application approval. New Drug Applications (NDA or innovator) are “N”. Abbreviated New Drug Applications (ANDA or generic) are “A”.
• appl_no: FDA assigned number to the application. Format is nnnnnn.
• product_no: FDA assigned number to identify the application products. Each strength is a separate product. May repeat for multiple part products. Format is nnn.
• exclcode: Code to designate exclusivity granted by the FDA to a drug product. Format is nnnnnnnnnn.
• excldatestr: The date the exclusivity expires. Format is MMM DD, YYYY.

Dataset 6: Ingredient

We load the ingredient dataset.

ingredient = read_csv("./data/ingredient.csv") %>%
  clean_names() %>%
  select(singnum, singredient, everything())

## Parsed with column specification:
## cols(
##   ingredient = col_character(),
##   singnum = col_integer(),
##   singredient = col_character()
## )

ingredient

## # A tibble: 3,371 x 3
##    singnum         singredient
##      <int>               <chr>
##  1       1    ABACAVIR SULFATE
##  2       1    ABACAVIR SULFATE
##  3       2 DOLUTEGRAVIR SODIUM
##  4       3          LAMIVUDINE
##  5       1    ABACAVIR SULFATE
##  6       2          LAMIVUDINE
##  7       1    ABACAVIR SULFATE
##  8       2          LAMIVUDINE
##  9       3          ZIDOVUDINE
## 10       1       ABALOPARATIDE
## # ... with 3,361 more rows, and 1 more variables: ingredient <chr>

The tidied ingredient dataset has 3371 observations and 3 variables:

• singnum: The number assigned to the single ingredient for each product.
• singredient: each single ingredient in the product.
• ingredient: The active ingredient(s) for the product. Multiple ingredients are in alphabetical order, separated by a semicolon.

Dataset 7: Drug by Condition

Scraping data on drugs per condition from this website

urls = paste("https://www.centerwatch.com/drug-information/fda-approved-drugs/medical-conditions/", toupper(letters), sep = "")

get_drug_disease = function(url) {
  
  disease_data = read_html(url) %>% 
  html_nodes(css = ".ToggleDrugCategory") %>%
  html_text() %>% 
  as_tibble() %>% 
  mutate(key = value) %>% 
  rename(disease = value)
  
  drug_data = read_html(url) %>% 
  html_nodes(css = ".CategoryListSection a:nth-child(1)") %>%
  html_text() %>% 
  as_tibble() %>% 
  mutate(key = value) %>% 
  rename(drug = value)
  
  drug_disease_data = read_html(url) %>% 
  html_nodes(css = "#MainColumn a:nth-child(1)") %>%
  html_text() %>% 
  as_tibble() %>% 
  mutate(key = value) %>% 
  rename(drug_disease = value)
  
  drug_disease_joined = left_join(drug_disease_data, drug_data, disease_data, by = "key") %>% 
  left_join(., disease_data, by = "key") %>%
  fill(disease) %>%
  select(-c(drug,drug_disease))
  
  toremove = drug_disease_joined %>% distinct(disease) %>% pull()
  
  drug_disease_joined = drug_disease_joined %>% 
  filter(!key %in% toremove) %>% 
  rename(drug = key) %>% 
  .[-c(1:29),]
  
  drug_disease_joined
}

drug_disease_data = map_df(urls, get_drug_disease) %>% 
  mutate(disease = str_to_title(disease),
         drug = str_to_title(drug)) %>% 
  rename(trade_name = drug) %>%
  mutate(trade_name = trimws(trade_name, which = c("both"))) %>% 
  separate(trade_name, into = c("trade_name", "remove"), sep = " ")

## Warning: Too many values at 959 locations: 2, 3, 4, 5, 6, 8, 9, 10, 11, 12,
## 13, 14, 16, 18, 19, 21, 23, 24, 25, 26, ...

## Warning: Too few values at 139 locations: 7, 64, 68, 76, 109, 122, 132,
## 144, 145, 161, 162, 173, 174, 188, 197, 207, 221, 223, 228, 231, ...

drug_disease_data

## # A tibble: 1,881 x 3
##     trade_name         remove disease
##  *       <chr>          <chr>   <chr>
##  1       Avita            Gel    Acne
##  2  Benzamycin  (Erythromycin    Acne
##  3 Clindamycin      Phosphate    Acne
##  4 Clindamycin      Phosphate    Acne
##  5    Differin     (Adapalene    Acne
##  6   Estrostep (Norethindrone    Acne
##  7     Finevin           <NA>    Acne
##  8      Klaron        (Sodium    Acne
##  9       Ortho     Tri-Cyclen    Acne
## 10     Retin-A          Micro    Acne
## # ... with 1,871 more rows

The drug_disease_data contains 1881 observations and 3 variables:

• drug: the drug name
• disease: the disease name

Combine Datasets into one

#A: join datasets by ingredient for ingredient and products (left join)
#products = longer dataset + ingredient
product_ingredient = left_join(products, ingredient, by = "ingredient")
product_ingredient

## # A tibble: 40,443 x 20
##                                           ingredient doseform route
##                                                <chr>    <chr> <chr>
##  1                                  ABACAVIR SULFATE SOLUTION  ORAL
##  2                                  ABACAVIR SULFATE SOLUTION  ORAL
##  3                                  ABACAVIR SULFATE   TABLET  ORAL
##  4                                  ABACAVIR SULFATE   TABLET  ORAL
##  5                                  ABACAVIR SULFATE   TABLET  ORAL
##  6                                  ABACAVIR SULFATE   TABLET  ORAL
##  7                                  ABACAVIR SULFATE   TABLET  ORAL
##  8                                  ABACAVIR SULFATE   TABLET  ORAL
##  9 ABACAVIR SULFATE; DOLUTEGRAVIR SODIUM; LAMIVUDINE   TABLET  ORAL
## 10 ABACAVIR SULFATE; DOLUTEGRAVIR SODIUM; LAMIVUDINE   TABLET  ORAL
## # ... with 40,433 more rows, and 17 more variables: trade_name <chr>,
## #   applicant <chr>, strength <chr>, drug_type <chr>, appl_no <int>,
## #   product_no <int>, tecodestr <chr>, tecode <int>, approval_month <chr>,
## #   approval_day <chr>, approval_year <chr>, otc <fctr>, appfull <chr>,
## #   ref_listed_drug <fctr>, ref_std <fctr>, singnum <int>,
## #   singredient <chr>

#B: exclusivity and patents left join (more~less)
patent_exclusivity = left_join(patents, exclusivity, by = c("appl_no", "product_no", "drug_type"))
patent_exclusivity

## # A tibble: 22,379 x 18
##    drug_type appl_no product_no patent_num pediatric_exclusivity_granted
##        <chr>   <int>      <int>      <chr>                        <fctr>
##  1         B   22425          1    5223510                            No
##  2         B   22425          1    5223510                            No
##  3         B   22307          1    5288726                            No
##  4         B   22307          1    5288726                            No
##  5         B   22307          1    5288726                           Yes
##  6         B   22307          1    5288726                           Yes
##  7         B   21462          1    5344932                            No
##  8         B   21462          1    5344932                           Yes
##  9         B   22088          1    5362718                            No
## 10         B   22088          1    5362718                           Yes
## # ... with 22,369 more rows, and 13 more variables:
## #   month_pat_expires <chr>, day_pat_expires <chr>,
## #   year_pat_expires <chr>, pat_use_code <chr>, delist <chr>,
## #   drug_substance_flag <fctr>, drug_prod_flag <fctr>,
## #   patent_delisted <fctr>, exclcode <chr>, excldatestr <chr>,
## #   month_excl_expires <chr>, date_excl_expires <chr>,
## #   year_excl_expires <chr>

#C: join A and B left join: "Orange Book." Thursday: meet to eliminate variables (inner join: only commonalities are preserved) #eliminate drugs that do not match
orange_book = left_join(product_ingredient, patent_exclusivity, by = c("appl_no", "product_no", "drug_type"))

orange_book = orange_book %>%
  mutate(ingredient = str_to_title(ingredient),
         doseform = str_to_title(doseform),
         route = str_to_title(route),
         trade_name = str_to_title(trade_name),
         applicant = str_to_title(applicant),
         otc = str_to_title(otc),
         appfull = str_to_title(appfull),
         singredient = str_to_title(singredient),
         strength = str_to_title(strength))
orange_book

## # A tibble: 64,423 x 35
##          ingredient doseform route       trade_name            applicant
##               <chr>    <chr> <chr>            <chr>                <chr>
##  1 Abacavir Sulfate Solution  Oral Abacavir Sulfate  Hetero Labs Ltd Iii
##  2 Abacavir Sulfate Solution  Oral           Ziagen        Viiv Hlthcare
##  3 Abacavir Sulfate Solution  Oral           Ziagen        Viiv Hlthcare
##  4 Abacavir Sulfate Solution  Oral           Ziagen        Viiv Hlthcare
##  5 Abacavir Sulfate Solution  Oral           Ziagen        Viiv Hlthcare
##  6 Abacavir Sulfate   Tablet  Oral Abacavir Sulfate           Apotex Inc
##  7 Abacavir Sulfate   Tablet  Oral Abacavir Sulfate Aurobindo Pharma Ltd
##  8 Abacavir Sulfate   Tablet  Oral Abacavir Sulfate  Hetero Labs Ltd Iii
##  9 Abacavir Sulfate   Tablet  Oral Abacavir Sulfate     Mylan Pharms Inc
## 10 Abacavir Sulfate   Tablet  Oral Abacavir Sulfate       Strides Pharma
## # ... with 64,413 more rows, and 30 more variables: strength <chr>,
## #   drug_type <chr>, appl_no <int>, product_no <int>, tecodestr <chr>,
## #   tecode <int>, approval_month <chr>, approval_day <chr>,
## #   approval_year <chr>, otc <chr>, appfull <chr>, ref_listed_drug <fctr>,
## #   ref_std <fctr>, singnum <int>, singredient <chr>, patent_num <chr>,
## #   pediatric_exclusivity_granted <fctr>, month_pat_expires <chr>,
## #   day_pat_expires <chr>, year_pat_expires <chr>, pat_use_code <chr>,
## #   delist <chr>, drug_substance_flag <fctr>, drug_prod_flag <fctr>,
## #   patent_delisted <fctr>, exclcode <chr>, excldatestr <chr>,
## #   month_excl_expires <chr>, date_excl_expires <chr>,
## #   year_excl_expires <chr>

# final goal: disease, funding, drug prices, and Orange Book

# eliminate unnecessary variables from orange book dataset
orange_book_clean = orange_book %>%
  rename(application_no = appl_no, 
         active_ingredient = singredient, 
         dose_form = doseform) %>%
  select(-c(ingredient, applicant, product_no, tecodestr, tecode, approval_day, ref_std, singnum, day_pat_expires, pat_use_code, drug_substance_flag, drug_prod_flag, exclcode, excldatestr, date_excl_expires)) %>% 
  select(trade_name, everything()) %>% 
  separate(dose_form, into = c("dose_form", "remove")) %>%
  filter(is.na(remove)) %>% 
  select(-remove) %>% 
  filter(str_detect(dose_form, regex("Tablet", ignore_case = TRUE))) %>%
  mutate(trade_name = trimws(trade_name, which = c("both")),
         trade_name = str_replace_all(trade_name, ", ", "-"),
         trade_name = str_replace_all(trade_name, " And ", "-"))

## Warning: Too many values at 8293 locations: 78, 79, 80, 81, 82, 153, 154,
## 155, 156, 157, 158, 159, 634, 635, 636, 1085, 1086, 1087, 1088, 1089, ...

## Warning: Too few values at 51953 locations: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
## 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, ...

# inner join of disease drug and orange book =y by trade_name
final_dataset =
  new_drug_price %>%
  inner_join(., orange_book_clean, by = c("trade_name", "drug_type", "otc")) %>% 
  distinct() %>% 
  left_join(., drug_disease_data, by = "trade_name") %>% 
  distinct() %>% 
  separate(disease, into = c("disease", "remove"), sep = " ") %>% 
  left_join(., final_NIH_funding, by = c("disease", "year")) %>% 
  as_tibble() %>% 
  select(-c(remove, remove.x)) %>%
  distinct()

## Warning: Column `otc` joining factor and character vector, coercing into
## character vector

## Warning: Too many values at 11500 locations: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
## 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, ...

## Warning: Too few values at 4247 locations: 1366, 1368, 1370, 1372, 1374,
## 1376, 1459, 1460, 1462, 1463, 1465, 1466, 1468, 1469, 1471, 1472, 1474,
## 1475, 1477, 1478, ...

Josie

# Remomve some unecessary datasets
rm(patent_exclusivity, product_ingredient)

Manali

#do rare diseases recieve less funding?
#set.seed(1)
#final_dataset %>%
#  sample_n(50)%>%
#plot_ly(x~funding, y~prevalence, type = "scatter")

#does funding correlate with drug prices?

#final_dataset %>%
#  sample_n(50)%>%
#  plot_ly(x ~ price, y ~ funding, type = "scatter")

Muhire

We created a spaghetti plot of the average price of drugs per year for each drug. We only used data on prices per drug unit (mainly per tablet). Information about the disease that the drugs treat can be seen by hovering over the plot. We notice that of the drugs with information on price per unit drug, the drugs for Hepatitis seem to have the highest cost. Unfortunately we cannot tell which kind of hepatitis these drugs treat.

Possible case studies on most popular (perceived) drugs: Lipitor, Xanax, Celebrex, Prozac, Crestor, Nexium, Lisinopril, Zoloft

final_dataset %>% 
  filter(pricing_unit == "EA" & !is.na(disease)) %>% 
  plot_ly(x = ~year, y = ~avg_price, type = "scatter", mode = "lines", text = ~paste("Drug: ", trade_name, '<br>Disease:', disease), color = ~trade_name) %>% 
  layout(showlegend = FALSE)

## Warning in RColorBrewer::brewer.pal(N, "Set2"): n too large, allowed maximum for palette Set2 is 8
## Returning the palette you asked for with that many colors

We also created a scatter plot of the mean NIH funding per drug (assumed to be the same as the NIH funding for disease research) against the demand of the drug (defined as the 2015 disease prevalence) colored by disease. Among drugs with available information on price per unit drug, obesity seems to have received the most funding on average and to have the highest 2015 prevalence.

final_dataset %>% 
  filter(pricing_unit == "EA" & !is.na(funding) & !is.na(prevalence)) %>% 
  group_by(trade_name, disease, prevalence) %>% 
  summarize(mean_fund = mean(funding)) %>% 
  plot_ly(x = ~prevalence, y = ~mean_fund, type = "scatter", alpha = 1, text = ~paste("Drug: ", trade_name, '<br>Disease:', disease), color = ~disease) %>%
  layout(showlegend = FALSE)

## No scatter mode specifed:
##   Setting the mode to markers
##   Read more about this attribute -> https://plot.ly/r/reference/#scatter-mode

## Warning in RColorBrewer::brewer.pal(N, "Set2"): n too large, allowed maximum for palette Set2 is 8
## Returning the palette you asked for with that many colors

MeOak